Fix incorrect envelope time conversions in disting EX, 1010music and Maschine 2/3

[douglas-carmichael]

While auditing the creators/detectors for envelope-time conversion bugs I found three independent issues. In each case the amplitude envelope of the same format already does it correctly, so the fix is to make the affected path match that reference. Each fix is small and independent.

1. disting EX — dropped hold time in amplitude decay

DistingExCreator writes the amplitude decay as:

Math.log (Math.max (0, envelope.getHoldTime ()) + Math.max (0, envelope.getDecayTime ()) / 0.02) / 0.0521

Operator precedence makes / 0.02 bind to the decay term only, i.e. log(hold + decay/0.02) instead of the intended log((hold + decay)/0.02). Because decay/0.02 dwarfs hold, the hold time is effectively dropped from the written decay. The reader confirms the intent — it decodes with decay = 0.02 * Math.exp (0.0521 * value), whose inverse is log((hold+decay)/0.02)/0.0521. Fixed by parenthesising (hold + decay) before the division.

2. 1010music (blackbox and Bento) — decay/release write scale doesn't match read scale

The amplitude decay and release are written with a 25 s full-scale but read back with a 38 s full-scale:

Parameter	Write (normalizeTime…)	Read (denormalizeTime)
Attack	9.0	9.0 ✅ consistent
Decay	25.0	38.0 ❌ mismatch
Release	25.0	38.0 ❌ mismatch

normalizeTime/denormalizeTime are inverse operations, so the same full-scale constant must be used on both sides or the value doesn't round-trip — a converted blackbox/Bento preset plays its decay and release noticeably shorter than the source. Attack is already consistent at 9 s. The reader's 38 s is the device-calibrated value (it was introduced when read-back of the amp envelope was implemented), so I aligned the writer (Music1010Creator and BentoCreator) to 38 s.

3. Maschine 2/3 — modulation (filter/pitch) envelope time conversions

In MaschinePresetAccessor the modulation envelope (used for the filter and pitch envelopes) is converted differently from the amplitude envelope in the same file:

• Reading: attack and decay are left in milliseconds (mapToAttackMillis / mapToDecayAndRelease without the / 1000.0 the amp envelope applies), so they come out a thousand times too long; the release skips mapToDecayAndRelease entirely and uses the raw normalized value.
• Writing: the release skips decayAndReleaseMillisToInput, writing raw milliseconds into the parameter row instead of the device's encoded value.

Fixed so the modulation envelope uses the same conversion as the amplitude envelope (read: … / 1000.0; write: decayAndReleaseMillisToInput (…)). Read and write offsets (148/156/160/164) already match, so this only corrects the value mapping.

---

All four changed files compile. (The tree currently has an unrelated XMLUtils.escapeAttribute compile error in OmnisphereCreator at main HEAD against the vendored uitools jar; it is independent of these changes.) CHANGELOG updated with a per-format entry.